Korean AI firms expand reach: Dinotisia joins AI Alliance, DeepX gets innovation nod, Cleon advances human interface

Dinotisia, a specialist in long-term memory artificial intelligence (AI) and integrated semiconductor solutions, has joined Kolon Benit's 'AI Alliance.' The company announced on August 18 that its inclusion as an AI specialized technology provider aims to expand AI construction projects across various industrial sectors.

The AI Alliance, operated by Kolon Benit, functions as an AI commercialization support platform. It connects participating companies' technologies with actual customer projects to generate commercialization outcomes. Since its launch in June 2024, it has grown into South Korea's largest AI collaborative ecosystem, with approximately 100 AI technology and solution companies participating.

Dinotisia leverages its expertise in enterprise AI construction, including vector databases, data accelerators, AI agents, and large/small language models (LLM/SLM). The company has experience building RAG-based AI briefing services for accounting firms, AI agent portals for aerospace and defense companies, and UX research services for automotive manufacturers.

Han Kwang-shin, Executive Managing Director at Dinotisia, stated, "This participation in the AI Alliance will expand our AI construction experience accumulated in accounting, legal, defense, aerospace, and manufacturing to more corporate sites." He added, "We will connect Kolon Benit's customer and partner network with Dinotisia's AI data infrastructure and agent technology to transform corporate data into tangible work performance."

“

This participation in the AI Alliance will expand our AI construction experience accumulated in accounting, legal, defense, aerospace, and manufacturing to more corporate sites. We will connect Kolon Benit's customer and partner network with Dinotisia's AI data infrastructure and agent technology to transform corporate data into tangible work performance.

In other AI news, DeepX, a company focused on ultra-low-power on-device AI semiconductors, announced on August 18 that its AI semiconductor module was designated an 'Excellent R&D Innovative Product' by the Ministry of Science and ICT. The designated products, the NPU-based AI inference modules 'DX-M1 and DX-H1 series,' enable real-time inference for AI models. This designation allows for three years of direct contracts with public institutions and provides procurement immunity for buyers in cases of intentional or gross negligence.

DeepX attributes its power efficiency to a proprietary 'Data Reuse' architecture that minimizes data movement during computation. According to the company, its DX-M1 based NPU module achieved up to a 93% reduction in power consumption compared to NVIDIA GPUs in official tests while maintaining equivalent high-resolution AI object detection performance. The company is expanding its presence in the defense and public sectors, having recently supplied its NPUs for an AI-based intelligent video surveillance system at an Air Force base.

Kim Nok-won, CEO of DeepX, expressed his ambition to "expand the application of domestic AI semiconductors in public infrastructure and create sovereign AI cases that prove economic feasibility in real-world scenarios." Sovereign AI refers to AI operation methods that ensure data sovereignty and independence.

“

This designation as an excellent R&D innovative product will expand the application of domestic AI semiconductors in public infrastructure and create sovereign AI cases that prove economic feasibility in real-world scenarios.

Furthermore, Cleon, a human-AI interface company, announced on August 19 that it is advancing its technology for AI to interact naturally with humans through faces, voices, expressions, and gestures. Their research has been accepted into the British Machine Vision Conference (BMVC) 2026, marking their third international conference presentation this year, following ICLR and CVPR. Cleon's technology combines LLMs, Automatic Speech Recognition (ASR), Text-to-Speech (TTS), RAG, and real-time video generation and streaming to create interfaces where AI can speak, emote, and react with gestures.

The BMVC 2026 paper focuses on modulating facial expression intensity and changes based on emotion. Previous papers presented at CVPR 2026 dealt with full-body conversational avatar technology that generates hand and body gestures from voice alone, while the ICLR 2026 paper addressed the expression of secondary details like clothing movement based on body motion. These technologies are implemented in products such as the conversational AI agent 'Rollfit' and the digital human development tool 'Cleon Studio SDK,' finding applications in corporate training, digital signage, and brand events.

Kang Ji-soo, CTO of Cleon, emphasized the need for AI to go beyond understanding and generating responses. "For AI to serve as a trainer or responder in front of actual people, simply understanding questions and generating answers is not enough," he said. "We will enhance the performance of individual technologies and connect them into a single real-time system to create interfaces where AI can interact more naturally with people."

Finally, Angel Robotics, a company specializing in wearable robots, held an international symposium on robot rehabilitation in Ho Chi Minh City, Vietnam, on August 14. The event, themed 'Application of Robotics in Rehabilitation Medicine: Trends and Implementation in Vietnam,' brought together experts from South Korea, Italy, and Vietnam to discuss the practical application and expansion of robot technology in rehabilitation medicine.

“

For AI to serve as a trainer or responder in front of actual people, simply understanding questions and generating answers is not enough. We will enhance the performance of individual technologies and connect them into a single real-time system to create interfaces where AI can interact more naturally with people.